Latency-Response Theory Model: Evaluating Large Language Models via Response Accuracy and Chain-of-Thought Length